Supplementary Figure 6: Correcting for the effect of sequencing coverage on the ascertainment of common primate variants.
From: Predicting the clinical impact of human mutation with deep neural networks
The probability of observing a given variant in a nonhuman primate species is inversely correlated with the sequencing depth at that position in the ExAC/gnomAD exome dataset. In contrast, the lower gnomAD read depth did not affect the probability of observing a common human variant at that position (>0.1% allele frequency) because the large number of human exomes sequenced makes ascertainment of common variation almost guaranteed. When picking matched variants for each of the primate variants for training the network, the probability of picking a variant was adjusted for the effects of sequencing depth, in addition to matching for trinucleotide context to control for mutational rate and gene conversion.
